基于离心泵数字孪生流场云图的叶轮故障诊断方法与应用

李亚洁; 刘强; 李炜

doi:10.13700/j.bh.1001-5965.2022.0997

基于离心泵数字孪生流场云图的叶轮故障诊断方法与应用

doi: 10.13700/j.bh.1001-5965.2022.0997

李亚洁^{1, 2, 3},
刘强^{1, 2, 3},
李炜^{1, 2, 3, ,}

1.
兰州理工大学电气工程与信息工程学院，兰州 730050
2.
兰州理工大学甘肃省工业过程先进控制重点实验室，兰州 730050
3.
兰州理工大学电气与控制工程国家级实验教学示范中心，兰州 730050

基金项目:

国家自然科学基金(62163022)；国家重点研发计划(2020YFB1713600)

详细信息

通讯作者:
E-mail：liwei@lut.edu.cn

中图分类号: TH165⁺.3；TP181
计量
- 文章访问数: 1066
- HTML全文浏览量: 269
- PDF下载量: 101
- 被引次数: 0
出版历程
- 收稿日期: 2022-12-17
- 录用日期: 2023-03-26
- 网络出版日期: 2023-04-20
- 整期出版日期: 2025-01-31

Impeller fault diagnosis method and application based on digital twin flow field contour of centrifugal pump

LI Yajie^{1, 2, 3},
LIU Qiang^{1, 2, 3},
LI Wei^{1, 2, 3
, ,}

1.
College of Electrical and Information Engineering，Lanzhou University of Technology，Lanzhou 730050，China
2.
Key Laboratory of Gansu Advanced Control for Industrial Processes，Lanzhou University of Technology，Lanzhou 730050，China
3.
National Demonstration Center for Experimental Electrical and Control Engineering Education，Lanzhou University of Technology，Lanzhou 730050，China

Funds:

National Natural Science Foundation of China (62163022)；National Key Research and Development Program of China (2020YFB1713600)

More Information

Corresponding author: E-mail：liwei@lut.edu.cn

摘要

摘要:
随着工业技术的发展，离心泵的健康诊断与维护需求日益迫切，结合数字孪生和机器视觉技术，提出一种基于数字孪生流场云图的离心泵叶轮机械故障智能诊断方法。借助离心泵数字孪生模型来模拟叶轮叶片随机断裂故障的演化发展，生成具有不同故障特征的叶轮流场压力及速度云图；基于对Yolov5算法的学习训练，获得了压力和速度云图两类机器视觉模型，并结合统计分析实现了叶轮故障的初步诊断；进而考虑两类检测模型的优势互补特性，基于堆叠集成的思想将二者融合，以提升叶轮故障诊断的准确性。经实验验证，针对叶轮叶片的随机断裂故障，所提方法可达到0.99以上的诊断准确度，开发的离心泵叶轮机械故障智能诊断系统使所提方法得以落地应用。
- 离心泵 /
- 数字孪生 /
- 叶轮机械 /
- 机器视觉 /
- 智能诊断
Abstract:
With the development of industrial technology, the health diagnosis and maintenance of centrifugal pumps are increasingly urgent. Combining digital twin and machine vision technology, this paper proposed an intelligent impeller fault diagnosis method for centrifugal pumps based on a digital twin flow field cloud diagram. First of all, the digital twin model of the centrifugal pump was used to simulate the evolution of the random fracture for the impeller blades, and the pressure and velocity cloud diagrams of the impeller flow field with different fault characteristics were generated. Secondly, based on the learning and training of the Yolov5 algorithm, two kinds of machine vision models, namely pressure and velocity cloud diagrams, were obtained, and the preliminary diagnosis of impeller fault was realized by combining statistical analysis. Furthermore, the complementary advantages of the two types of detection models were considered, and the two types of detection models were combined based on the idea of stack integration to improve the accuracy of impeller fault diagnosis. The experimental verification shows that the intelligent fault diagnosis method for centrifugal pumps proposed in this paper has a diagnosis accuracy of more than 0.99 for the random fracture of impeller blades. The developed intelligent impeller fault diagnosis system for centrifugal pumps makes the method developed in this paper be applied to practical scenarios.
- centrifugal pump /
- digital twins /
- impeller /
- machine vision /
- intelligent diagnosis

HTML全文

图 1 叶轮机械故障诊断方案

Figure 1. Impeller fault diagnosis scheme

下载: 全尺寸图片幻灯片

图 2 叶轮机械故障智能诊断方法

Figure 2. Intelligent impeller fault diagnosis method

下载: 全尺寸图片幻灯片

图 3 检测结果对比

Figure 3. Comparison of detection results

下载: 全尺寸图片幻灯片

图 4 离心泵叶轮机械故障智能诊断系统的核心功能界面

Figure 4. Core function interface of intelligent impeller fault diagnosis system for centrifugal pump

下载: 全尺寸图片幻灯片

表 1 Yolov5故障检测模型训练参数设置

Table 1. Setting of training parameters for Yolov5 fault detection model

网络参数	数值
迭代次数	300
批次大小	8
动量因子	0.9
学习速率	0.001
权重衰减系数	0.0005
置信度阈值	0.5
非极大抑制阈值	0.3

下载: 导出CSV

表 2 模型损失

Table 2. Model loss

检测模型	损失名称	训练集损失	测试集损失
压力云图检测模型	定位损失	0.0513	0.0435
	置信损失	0.0225	0.0156
	分类损失	0.0031	0.0014
速度云图检测模型	定位损失	0.0514	0.0442
	置信损失	0.0205	0.0151
	分类损失	0.0032	0.0015

下载: 导出CSV

表 3 模型性能指标

Table 3. Model performance indicators

检测模型	P_recision	R_ecall	A_pre	m_AP@0.5
压力云图检测模型	0.940	0.939	0.949	0.948
速度云图检测模型	0.976	0.956	0.967	0.975

下载: 导出CSV

表 4 Resnet模型性能

Table 4. Resnet model performance

检测模型	训练集损失	训练集准确度	测试集损失	测试集准确度
压力云图识别模型	0.3317	0.6792	0.5236	0.6528
速度云图识别模型	0.7396	0.5088	0.7701	0.5001

下载: 导出CSV

表 5 融合准确度对比

Table 5. Fusion accuracy comparison

融合方式	准确度
加权融合	0.9917
K近邻	0.9941
逻辑回归	0.9901
随机森林	0.9938
决策树	0.9946
朴素贝叶斯	0.9925

下载: 导出CSV

参考文献(32)

[1]	TONG Z M, XIN J G, TONG S G, et al. Internal flow structure, fault detection, and performance optimization of centrifugal pumps[J]. Journal of Zhejiang University Science A: Applied Physics & Engineering, 2020, 21(2): 85-117.
[2]	周耀. 离心泵故障诊断研究现状与发展趋势[J]. 设备管理与维修, 2019(20): 161-163. ZHOU Y. Research status and development trend of centrifugal pump fault diagnosis[J]. Plant Maintenance Engineering, 2019(20): 161-163(in Chinese).
[3]	唐一科, 柯研, 谢志江. 叶轮机械叶片状态监测与故障诊断的现状与发展[J]. 噪声与振动控制, 2003, 23(6): 5-8. TANG Y K, KE Y, XIE Z J. The actuality and development of turbo machine’s status monitoring and fault diagnose[J]. Noise and Vibration Control, 2003, 23(6): 5-8(in Chinese).
[4]	唐亮, 吴桐, 杜鹏, 等. 基于激光条带追踪的桥梁振动非接触测量方法[J]. 工程科学与技术, 2022, 54(6): 135-145. TANG L, WU T, DU P, et al. Non-contact measurement method of bridge vibration based on laser strip tracing[J]. Engineering Science and Technology, 2022, 54(6): 135-145.
[5]	STASZEWSKI W J, BIN JENAL R, KLEPKA A, et al. A review of laser Doppler vibrometry for structural health monitoring applications[J]. Key Engineering Materials, 2012, 518: 1-15.
[6]	FROULA D H. Plasma scattering of electromagnetic radiation theory and measurement techniques[M]. Burlington: Academic Press/Elsevier, 2011.
[7]	ANDRENELLI L, PAONE N, ROSSI G L. Large-bandwidth reflection fiber-optic sensors for turbomachinery rotor blade diagnostics[J]. Sensors and Actuators A: Physical, 1992, 32(1-3): 539-542.
[8]	CAETANO E, SILVA S, BATEIRA J. A vision system for vibration monitoring of civil engineering structures[J]. Experimental Techniques, 2011, 35(4): 74-82.
[9]	PFISTER T, GÜNTHER P, DREIER F, et al. Novel dynamic rotor and blade deformation and vibration monitoring technique[J]. Journal of Engineering for Gas Turbines and Power, 2012, 134(1): 1.
[10]	WEISS A, GEISLER R, MULLER M M, et al. Dynamic-stall measurements using time-resolved pressure-sensitive paint on double-swept rotor blades[J]. Experiments in Fluids, 2021, 63(1): 15.
[11]	黄瑾珉. 面向机械故障智能诊断神经网络优化与识别方法研究[J]. 化学工程与装备, 2019(4): 220-224. HUANG J M. Research on optimization and recognition method of neural network for intelligent diagnosis of mechanical faults[J]. Chemical Engineering and Equipment, 2019(4): 220-224(in Chinese).
[12]	ALI SAUD ALTOBI M, BEVAN G, WALLACE P, et al. Fault diagnosis of a centrifugal pump using MLP-GABP and SVM with CWT[J]. Engineering Science and Technology, an International Journal, 2019, 22(3): 854-861.
[13]	PROSVIRIN A E, AHMAD Z, KIM J M. Global and local feature extraction using a convolutional autoencoder and neural networks for diagnosing centrifugal pump mechanical faults[J]. IEEE Access, 2021, 9: 65838-65854.
[14]	CHEN Z H, CEN J, XIONG J B. Rolling bearing fault diagnosis using time-frequency analysis and deep transfer convolutional neural network[J]. IEEE Access, 2020, 8: 150248-150261.
[15]	KUMAR A, GANDHI C P, ZHOU Y Q, et al. Improved deep convolution neural network (CNN) for the identification of defects in the centrifugal pump using acoustic images[J]. Applied Acoustics, 2020, 167(10): 107399.
[16]	HASAN M J, RAI A, AHMAD Z, et al. A fault diagnosis framework for centrifugal pumps by scalogram-based imaging and deep learning[J]. IEEE Access, 2021, 9: 58052-58066.
[17]	ZHANG Y, HUANG X B, JIA J Y, et al. A recognition technology of transmission lines conductor break and surface damage based on aerial image[J]. IEEE Access, 2019, 7: 59022-59036.
[18]	CHENG L B, LI J, DUAN P, et al. A small attentional YOLO model for landslide detection from satellite remote sensing images[J]. Landslides, 2021, 18(8): 2751-2765.
[19]	American Petroleum Institute. Centrifugal pumps for the petroleum, petrochemical and natural gas industries, 12th edition: API 610[S]. United States: American Petroleum Institute, 2021: 68-80.
[20]	王燕, 马文文. 矿用通风机叶片裂纹诊断方法的研究[J]. 同煤科技, 2018(6): 24-26. WANG Y, MA W W. Research on diagnosis method of fan blade crack in coal mine[J]. Datong Coal Science & Technology, 2018(6): 24-26(in Chinese).
[21]	陈雪巍, 汪轶俊, 何岗, 等. 基于CFD的某运载火箭发动机燃料泵故障诊断分析[J]. 上海航天, 2016, 33(2): 68-72. CHEN X W, WANG Y J, HE G, et al. Faulty analysis of a centrifugal pump for some liquid rocket engine based on CFD[J]. Aerospace Shanghai (Chinese & English), 2016, 33(2): 68-72(in Chinese).
[22]	MADNI A M, MADNI C C, LUCERO S D. Leveraging digital twin technology in model-based systems engineering[J]. Systems, 2019, 7(1): 7.
[23]	XIE R, CHEN M Y, LIU W H, et al. Digital twin technologies for turbomachinery in a life cycle perspective: a review[J]. Sustainability, 2021, 13(5): 2495.
[24]	张胜文, 杨凌翮, 程德俊. 数字孪生驱动的离心泵机组故障诊断方法研究[J]. 计算机集成制造系统, 2023(5): 1462-1470. ZHANG S W, YANG L H, CHENG D J. Research on fault diagnosis method of centrifugal pump unit driven by digital twin[J]. Computer Integrated Manufacturing System, 2023(5): 1462-1470(in Chinese).
[25]	YUSUPBEKOV N, ADILOV F, IVANYAN A. Development of digital twin for centrifugal rotating equipment assets[M]//Lecture Notes in Networks and Systems. Cham: Springer International Publishing, 2022: 446-455.
[26]	LU Q C, PARLIKAD A K, WOODALL P, et al. Developing a digital twin at building and city levels: Case study of West Cambridge campus[J]. Journal of Management in Engineering, 2020, 36(3): 05020004.
[27]	AENIS M, KNOPF E, NORDMANN R. Active magnetic bearings for the identification and fault diagnosis in turbomachinery[J]. Mechatronics, 2002, 12(8): 1011-1021.
[28]	冯黎刚. 相对定位法检测 GNSS 相位中心偏差的研究[J]. 陕西交通科教研究, 2021(2): 5-9. FENG L G. Study on detection of GNSS phase center deviation by relative positioning method[J]. Shanxi Transportation Science and Education Research, 2021(2): 5-9(in Chinese).
[29]	胡振江. 单级单吸式离心泵叶片裂纹故障诊断技术研究[D]. 武汉: 华中科技大学, 2019: 15-20. HU Z J. Research on fault diagnosis technology of single-stage single-suction centrifugal pump vane crack[D]. Wuhan: Huazhong University of Science and Technology, 2019: 15-20(in Chinese).
[30]	AHMAD Z, PROSVIRIN A E, KIM J M, et al. Multistage centrifugal pump fault diagnosis by selecting fault characteristic modes of vibration and using Pearson linear discriminant analysis[J]. IEEE Access, 2020, 8: 223030-223040.
[31]	WANG D D, HE D J. Channel pruned YOLO V5s-based deep learning approach for rapid and accurate apple fruitlet detection before fruit thinning[J]. Biosystems Engineering, 2021, 210: 271-281.
[32]	XU Z B, HUANG X P, HUANG Y, et al. A real-time zanthoxylum target detection method for an intelligent picking robot under a complex background, based on an improved YOLOv5s architecture[J]. Sensors, 2022, 22(2): 682.